A transparent rubric that you can audit, argue with, or use to score tools yourself. Every review on this site applies the same framework — same categories, same weights, same benchmarks — regardless of who makes the tool or what they've offered us.
Every tool on Dappiehub gets a score in each of five universal categories. The category scores are weighted and combined to produce the headline 0–10 score you see on tool cards. The five categories were chosen because they're the ones that consistently determine whether a piece of business software is worth paying for.
Time to first useful output. How quickly a non-technical owner can install the tool, complete a real task with it, and get a result they'd actually use.
Whether the tool delivers on its core promise meaningfully better than free alternatives — or just adds polish to something a generic tool already does.
What you get per pound spent at each pricing tier — and whether the cheapest plan is actually usable for real work, or a marketing tease.
How well the tool connects to the rest of a small business stack — Word, Google Drive, common CRMs, automation tools, payment platforms.
Uptime, support quality, data security posture, and whether the company looks likely to still be here in two years. New AI startups die fast.
The five universal categories give us a like-for-like score across every tool. But comparing an AI writing tool against an AI image generator on "Feature Depth" means little if you don't know what features matter for that category. So every tool also gets scored against a category-specific rubric.
The 14 families currently covered are:
Within each family, the rubric weights are tuned to that category. A coding tool's "feature depth" score should reward IDE integration and autonomy; an image tool's should reward prompt fidelity and style range. The full per-family rubric appears on each individual tool page under "Category Scores".
Numerical scores in a vacuum mean little. A "9.1" only matters if you know what an 8.0 looks like. To prevent score inflation, every tool we review is calibrated against two reference benchmarks that don't move:
The world's most widely-used AI tool, scored as it stood at our most recent review cycle. Any tool claiming to be a better generalist needs to beat it on something specific to earn a higher score.
The reference for output quality and reasoning depth, particularly on long-form and document work. A tool scoring above Claude on quality categories needs to demonstrably outperform it.
This calibration is why our headline scores cluster between 7.5 and 9.5 rather than spanning the full 0–10 range. Tools below 7.0 generally aren't included in the catalogue at all — there's no point reviewing software we wouldn't recommend to anyone. The tools we cover are tools we think have a legitimate case for being chosen by someone.
A new review is roughly a two-week process. The steps are the same regardless of whether the tool is new on the market or a long-established player:
We use the tool on actual work — not vendor demos — for at least a week. Writing real emails, drafting real proposals, processing real documents, automating real workflows. Marketing screenshots are not evidence.
We try the free tier (if there is one) and at least the next paid tier up. Tools whose free tier is unusable get marked down even if their paid tier is excellent — most readers will never get past the free tier to find out.
We test the tool against a representative small-business stack — Microsoft 365, Google Workspace, Slack or similar, a common CRM, and Zapier or Make. Tools that don't play well with the existing world get marked down.
Direct head-to-head against ChatGPT and Claude on the same tasks where relevant. If a generalist tool can do the same job for free, the specialist tool needs a real reason to exist.
Scores are assigned to all five universal categories and the relevant family rubric. The verdict, pros, cons and "best for" recommendations are written. The review is published with a clear date, and the tool enters the rolling re-review cycle.
AI tools change month to month — pricing changes, models get upgraded, features get added or removed. A review written six months ago can be materially wrong today. To keep the catalogue honest, every tool is on a rolling re-review schedule:
Each tool page shows its Last Updated date prominently. If a date is older than 180 days when you read this, that's a bug we want to know about — please let us know.
Dappiehub does not currently take affiliate commission on tool sign-ups. The "Try X →" links on tool pages are direct links to the vendor, with no tracking or commission attached.
If this changes — and most independent review sites at our scale eventually do take affiliate links — the policy will be:
Sponsored content, where it exists, will be clearly labelled as "Sponsored" and held to the same factual accuracy standards as editorial content. Sponsored placement does not move scores.
We will get things wrong. Pricing changes we miss. New features we underweight. Tools we score generously and shouldn't have, or harshly and shouldn't have either. When this happens, the response is straightforward:
Spotted an error? Email corrections@dappiehub.com with the page URL and the issue. We respond within five working days, correct verifiable factual errors promptly, and note material corrections at the bottom of the affected page along with the date of correction.
Scoring disagreements aren't factual errors — they're editorial judgement, and we'll defend our scores even when readers (or vendors) disagree with them. But if we've got a feature wrong, missed a pricing tier, or misrepresented what a tool does, that's a correction and we'll make it.
Every tool review on Dappiehub uses the framework above. Browse the catalogue or take the finder for a personalised shortlist.